Abstract: Multi-energy load forecasting technology is the key cornerstone to ensure the supply and demand balance and stable operation of integrated energy system (IES). However, IES load with strong randomness and volatility aggravates the difficulty of accurate ultra short term multi-energy load forecast. Therefore, the optimal Bagging ensemble ultra short term multi-energy load forecasting method considering least average envelope entropy load decomposition is proposed. The parameters optimization model of variational mode decomposition based on least average envelope entropy is constructed, and the multi-energy load of IES is decomposed into the set of intrinsic mode functions; the strong correlation characteristic of calendar, weather and load of multi-energy load forecasting are filtered based on the uniform information coefficient method. Combined with the IMFs set of load, calendar rules, meteorological environment and load data, the Bagging ensemble ultra short term multi-energy load forecasting model is constructed, the ensemble strategy optimization model is constructed based on the mean absolute percentage error and R-square, and then the optimal ensemble strategy and final forecast results are also obtained. Simulation verification is carried out with IES of Arizona State University Tempe Campus as the object. The results show that the mean absolute percentage error of the proposed method in electric, heat and cooling load forecasting is 1.9486%, 2.0585% and 2.5331%, respectively, which has higher accuracy than other forecast methods.